CS395T Project 2: Shift-Reduce Parsing
|
|
- Flora Ray
- 5 years ago
- Views:
Transcription
1 CS395T Project 2: Shift-Reduce Parsing Due date: Tuesday, October 17 at 9:30am In this project you ll implement a shift-reduce parser. First you ll implement a greedy model, then you ll extend that model to be a global model using beam search, with appropriate structured training. Goal: The primary goal of this assignment is to expose you to different ways of doing approximate inference and learning in intractably large state spaces. You ll learn how to manage complex inferential state spaces at both training time and test time. Background Dependency trees are a form of tree-structured syntactic annotation. In this formalism, each word is assigned one parent word these parent assignments have to form a directed acyclic graph. The standard dataset used in English parsing is the Penn Treebank. This dataset is annotated with constituency syntax, but several different rule-based dependency conversions have been proposed; we use Stanford dependencies in this project. Here s an example of a sentence from the dataset: 1 Not _ ADV RB _ 3 neg 2 all _ DET PDT _ 3 predet 3 those _ DET DT _ 6 nsubj 4 who _ PRON WP _ 5 nsubj 5 wrote _ VERB VBD _ 3 rcmod 6 oppose _ VERB VBP _ 0 root 7 the _ DET DT _ 8 det 8 changes _ NOUN NNS _ 6 dobj 9. _ PUNCT. _ 6 punct Not all those who wrote oppose the changes. Figure 1: Dependency parse. Arrows are drawn from parents to children. The columns are (1) sentence index; (2) word; (3) blank; (4) coarse part-of-speech tag; (5) part-of-speech tag; (6) blank; (7) index (1-based) of syntactic parent; (8) dependency label; (9-10) blank. Dependency parsing is the task of predicting the syntactic parents (column 7) and dependency labels (column 8). In this project, we ll be doing unlabeled dependency parsing (unless you choose to explore labeled parsing), which just involves predicting the parents. Moreover, we will be assuming access to gold part of speech tags for feature extraction. 1 Systems for dependency parsing either operate in a graph-based or transition-based paradigm. Graphbased systems search over all trees using dynamic programs and are typically trained in structured logistic regression or max-margin frameworks, similar to CRFs for NER. Transition-based systems make local decisions and process the trees incrementally from left-to-right. In this project, you ll implement both a purely greedy transition-based system as well as a global transition-based system that approximately optimizes the sequence of transitions using beam search, with scores summed over all transitions. Building a tree from a sequence of local decisions involves a shift-reduce system similar to deterministic parsing algorithms used in compilers. The parser s state is captured by a stack of partial dependency trees 1 In practice, systems either run a part-of-speech tagger to produce predicted part-of-speech tags or do tagging as well as parsing, which can be implemented as an extension of the shift-reduce framework (Bohnet and Nivre, 2012). 1
2 and a buffer of words waiting to be processed. There are several related systems for modeling trees as decisions in this framework. We use the arc-standard framework; see Nivre (2004) or Huang et al. (2009) for more details on how this works. Code is provided to you that produces a transition sequence from a parse using the lowest stack depth heuristic: attach left children eagerly rather than keeping them on the stack until right children are finished. Dependency parsing is evaluated according to two heuristics: unlabeled attachment score (UAS) and labeled attachment score (LAS). UAS counts how many words in the evaluation set are assigned the correct parent, and divides this by the total number of words that is, it s simply the model s per-word accuracy at parent prediction. LAS is the same but also requires the dependency label to be correct; LAS is therefore always lower than UAS. In this project, you re only expected to do unlabeled parsing, so you can return a placeholder dependency label whenever one is required. Your Task In this project, you ll be building two related shift-reduce parsers: 1. A purely greedy parser that is implemented via a classifier over shift/reduce decisions 2. A global beam-search parser that searches over the same space but does inference over whole sequences You will also be extending your project in some way beyond this. Getting Started To start, try running python parser.py This simply synthesizes the sequence of transitions for each sentence in the development set and checks that these reconstruct the original tree they do nearly 100% of the time. The code in treedata.py loads the data into instances of ParsedSentence. This class contains both a list of Token objects similar to the representation used in Project 1, though this time Tokens contain the word, part-of-speech, and a coarse part-of-speech. It also contains a list of Dependency objects, indexed according to words in the sentence, with each object containing the index of the parent (0-based, -1 is the root) and the dependency label. You are provided with code that finds a canonical transition sequence given a syntactic parse. Specifically, the get decision sequence function takes a ParsedSentence object and returns two parallel lists of decisions ( S, L, or R ) and states (ParserState objects). This conversion follows the arc-standard transition system. ParserState objects consist of three data structures: the stack (a list of word indices, top of the stack last), the buffer (a list of word indices, next word first), and the current set of dependencies (a dictionary mapping child word indices to parent word indices). Several convenience methods are included to (a) access various parts of these data structures, (b) check conditions about them, and (c) execute transitions and return new ParserState objects. You should familiarize yourself with this representation of parser state. Part 1: Greedy Shift-Reduce Parsing In the first part of this project, you will implement a purely greedy version of a shift-reduce parser. This parser is a simple n-way classifier among decisions based on a current ParserState. The function 2
3 extract features provides a baseline feature set that looks at the top two words of the stack and the first few words of the buffer. GreedyModel is provided as a convenience class to store your code that can be run from the main parser.py Applying decisions to the parser state will lead to new parser states. Keep in mind that not every decision will be legal in every state. At the end of decoding, you can use the state.get dep objs method to convert the dependencies associated with a ParserState into Dependency objects suitable for evaluation. You are given more freedom here to decide how to build the greedy classifier: you can choose your classifier (perceptron, structured SVM, logistic regression), optimization techniques, training regimen, and features. The baseline feature set is a bit more impoverished than what was given to you for NER. You ll probably still want to use some optimizations like in Project 1, such as caching features and The minimum performance requirement on the development set here is 78 UAS. This is a pretty low bar and is achievable with the current feature set using SGD. Our reference implementation on this part got 86.8 UAS using Adagrad and additional features looking at children of items on the stack. With additional feature tuning (see (Zhang and Nivre, 2011) for some ideas of better features), you can get close to 90 UAS. Training on 1000 sentences, you should still get over 70, so you can develop features on a smaller dataset. Part 2: Global Beam Search and Structured Learning In this part, rather than scoring isolated classification decisions, you will instead be ranking parses according to the sum of the scores of all decisions made to create those parses. Training should now also take into account that we re doing a global search. You might want to follow these rough steps: 1. Implement beam search for global decoding. You should be able to run this with the greedy model and get decent results, though increasing the beam size might not cause things to be better (why not?) 2. Implement structured learning with either structured perceptron (non-loss augmented decode) or structured SVM (loss-augmented decode) for this purpose. You can either produce complete parses when computing your gradients or use early updating where you update as soon as the gold hypothesis falls off the beam. To assist you with beam search, we ve provided a Beam class in utils.py. This maintains parallel lists of elements and scores, sorted by score. It supports two operations: adding a scored item and returning the list of scored items. Only the k highest-scoring items are kept in the beam; adding an item that already exists in the beam will cause that items score to be max(new score, old score). Our implementation isn t particularly efficient, but it shouldn t be the bottleneck in your code, though you re free to optimize it further! Your beam search parser should be able to get at least 75 UAS depending on how well-tuned your greedy model is you may find it hard to beat your greedy model s performance. However, you should observe that increasing the beam size from 1 to higher values gives better results within your structured learning model. Running with a beam size more than 10 or 20 might be prohibitively slow, but see if you can get some small beam sizes (single digits) working well! Try training on smaller training sets for debugging if the full dataset still proves too slow. Implementation Tips One aspect you ll need to deal with is the fact that you ll encounter novel ParserState configurations during training, potentially leading to features you ve never seen before. Do you want to expand 3
4 the feature vector on-the-fly? Pre-extract a certain set of features (this could be the set of features from part 1) and ignore any features not in this set? How do you want to handle feature caching? Feel free to reuse learning code from Project 1 that you find useful! Part 3: Extensions These are suggestions for extensions. You also don t need to address each part of each section these are just suggestions, use your judgment about what you think constitutes a thorough enough extension. Training Explore additional tradeoffs between beam search and greedy search. What happens if you run the greedy parameters in the beam system at test time? Vice versa? Try a significantly larger or smaller feature set and see if the two parsers differ. Dynamic Oracle Implement the dynamic oracle of (Goldberg and Nivre, 2012). This is pretty ambitious! Modeling Implement a different transition systems, such as arc eager (Nivre, 2003; Nivre et al., 2006; Zhang and Clark, 2008), arc hybrid, or arc swift (Qi and Manning, 2017). How do the errors compare? You can also experiment with labeled dependency parsing. Features solve? Implement additional features from Zhang and Nivre (2011). What kinds of errors do these Other languages Experiment with dependency parsing in other languages. What s different about these languages? How do errors differ? Arabic, Chinese, and Japanese are included ask the instructor if there are other languages you re especially interested in. 2 Submission and Grading You should submit three files on Canvas: 1. Your code as a zip file or gzipped tar 2. Your best parser s output on the blind test set in CoNLL format. Set the appropriate flag in parser.py to produce this. 3. A report of around 2 pages, not including references, though you aren t expected to reference many papers. Your report should list your collaborators, briefly restate the core problem and what you are doing, describe relevant details of your implementation, present results from your greedy and beamed parsers as well, discuss these, describe your extension and give results for that, and optionally discuss error cases addressed by your extension or describe how the system could be further improved. Your report should be written in the tone and style of an ACL/NIPS conference paper. Any LaTeX format with reasonably small (1 margins) is fine, including the ACL style files 3 or any other one- or twocolumn format with equivalent density. 2 Treebanks available for Arabic, Basque, Bulgarian, Catalan, Chinese, Czech, Danish, Dutch, German, Greek, Hungarian, Italian, Japanese, Portuguese, Slovene, Spanish, Swedish, and Turkish. 3 Available at 4
5 Slip Days Per the syllabus, you have seven slip days to use during the semester, so you may choose to use part or all of your budget on this process to turn it in late as needed. References Bernd Bohnet and Joakim Nivre A Transition-Based System for Joint Part-of-Speech Tagging and Labeled Non-Projective Dependency Parsing. In Proceedings of the Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning (EMNLP-CoNLL). Yoav Goldberg and Joakim Nivre A Dynamic Oracle for Arc-Eager Dependency Parsing. In Proceedings of the International Conference on Computational Linguistics (COLING). Liang Huang, Wenbin Jiang, and Qun Liu Bilingually-constrained (Monolingual) Shift-reduce Parsing. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP). Joakim Nivre, Johan Hall, Jens Nilsson, Gülşen Eryiǧit, and Svetoslav Marinov Labeled Pseudo-projective Dependency Parsing with Support Vector Machines. In Proceedings of the Tenth Conference on Computational Natural Language Learning (CoNLL). Joakim Nivre An Efficient Algorithm for Projective Dependency Parsing. In Proceedings of the 8th International Workshop on Parsing Technologies (IWPT). Joakim Nivre Incrementality in Deterministic Dependency Parsing. In Proceedings of the Workshop on Incremental Parsing: Bringing Engineering and Cognition Together. Peng Qi and Christopher D. Manning Arc-swift: A Novel Transition System for Dependency Parsing. In Proceedings of the Association for Computational Linguistics (ACL). Yue Zhang and Stephen Clark A Tale of Two Parsers: Investigating and Combining Graph-based and Transition-based Dependency Parsing Using Beam-search. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP). Yue Zhang and Joakim Nivre Transition-based Dependency Parsing with Rich Non-local Features. In Proceedings of the Association for Computational Linguistics. 5
Introduction to Data-Driven Dependency Parsing
Introduction to Data-Driven Dependency Parsing Introductory Course, ESSLLI 2007 Ryan McDonald 1 Joakim Nivre 2 1 Google Inc., New York, USA E-mail: ryanmcd@google.com 2 Uppsala University and Växjö University,
More informationTransition-Based Dependency Parsing with Stack Long Short-Term Memory
Transition-Based Dependency Parsing with Stack Long Short-Term Memory Chris Dyer, Miguel Ballesteros, Wang Ling, Austin Matthews, Noah A. Smith Association for Computational Linguistics (ACL), 2015 Presented
More informationA Quick Guide to MaltParser Optimization
A Quick Guide to MaltParser Optimization Joakim Nivre Johan Hall 1 Introduction MaltParser is a system for data-driven dependency parsing, which can be used to induce a parsing model from treebank data
More informationTransition-based Dependency Parsing with Rich Non-local Features
Transition-based Dependency Parsing with Rich Non-local Features Yue Zhang University of Cambridge Computer Laboratory yue.zhang@cl.cam.ac.uk Joakim Nivre Uppsala University Department of Linguistics and
More informationAgenda for today. Homework questions, issues? Non-projective dependencies Spanning tree algorithm for non-projective parsing
Agenda for today Homework questions, issues? Non-projective dependencies Spanning tree algorithm for non-projective parsing 1 Projective vs non-projective dependencies If we extract dependencies from trees,
More informationFast(er) Exact Decoding and Global Training for Transition-Based Dependency Parsing via a Minimal Feature Set
Fast(er) Exact Decoding and Global Training for Transition-Based Dependency Parsing via a Minimal Feature Set Tianze Shi* Liang Huang Lillian Lee* * Cornell University Oregon State University O n 3 O n
More informationEasy-First POS Tagging and Dependency Parsing with Beam Search
Easy-First POS Tagging and Dependency Parsing with Beam Search Ji Ma JingboZhu Tong Xiao Nan Yang Natrual Language Processing Lab., Northeastern University, Shenyang, China MOE-MS Key Lab of MCC, University
More informationStack- propaga+on: Improved Representa+on Learning for Syntax
Stack- propaga+on: Improved Representa+on Learning for Syntax Yuan Zhang, David Weiss MIT, Google 1 Transi+on- based Neural Network Parser p(action configuration) So1max Hidden Embedding words labels POS
More informationHomework 2: Parsing and Machine Learning
Homework 2: Parsing and Machine Learning COMS W4705_001: Natural Language Processing Prof. Kathleen McKeown, Fall 2017 Due: Saturday, October 14th, 2017, 2:00 PM This assignment will consist of tasks in
More informationLet s get parsing! Each component processes the Doc object, then passes it on. doc.is_parsed attribute checks whether a Doc object has been parsed
Let s get parsing! SpaCy default model includes tagger, parser and entity recognizer nlp = spacy.load('en ) tells spacy to use "en" with ["tagger", "parser", "ner"] Each component processes the Doc object,
More informationTekniker för storskalig parsning: Dependensparsning 2
Tekniker för storskalig parsning: Dependensparsning 2 Joakim Nivre Uppsala Universitet Institutionen för lingvistik och filologi joakim.nivre@lingfil.uu.se Dependensparsning 2 1(45) Data-Driven Dependency
More informationNLP in practice, an example: Semantic Role Labeling
NLP in practice, an example: Semantic Role Labeling Anders Björkelund Lund University, Dept. of Computer Science anders.bjorkelund@cs.lth.se October 15, 2010 Anders Björkelund NLP in practice, an example:
More informationDynamic Feature Selection for Dependency Parsing
Dynamic Feature Selection for Dependency Parsing He He, Hal Daumé III and Jason Eisner EMNLP 2013, Seattle Structured Prediction in NLP Part-of-Speech Tagging Parsing N N V Det N Fruit flies like a banana
More informationTransition-Based Dependency Parsing with MaltParser
Transition-Based Dependency Parsing with MaltParser Joakim Nivre Uppsala University and Växjö University Transition-Based Dependency Parsing 1(13) Introduction Outline Goals of the workshop Transition-based
More informationarxiv: v1 [cs.cl] 25 Apr 2017
Joint POS Tagging and Dependency Parsing with Transition-based Neural Networks Liner Yang 1, Meishan Zhang 2, Yang Liu 1, Nan Yu 2, Maosong Sun 1, Guohong Fu 2 1 State Key Laboratory of Intelligent Technology
More informationDependency Parsing 2 CMSC 723 / LING 723 / INST 725. Marine Carpuat. Fig credits: Joakim Nivre, Dan Jurafsky & James Martin
Dependency Parsing 2 CMSC 723 / LING 723 / INST 725 Marine Carpuat Fig credits: Joakim Nivre, Dan Jurafsky & James Martin Dependency Parsing Formalizing dependency trees Transition-based dependency parsing
More informationHybrid Combination of Constituency and Dependency Trees into an Ensemble Dependency Parser
Hybrid Combination of Constituency and Dependency Trees into an Ensemble Dependency Parser Nathan David Green and Zdeněk Žabokrtský Charles University in Prague Institute of Formal and Applied Linguistics
More informationProjective Dependency Parsing with Perceptron
Projective Dependency Parsing with Perceptron Xavier Carreras, Mihai Surdeanu, and Lluís Màrquez Technical University of Catalonia {carreras,surdeanu,lluism}@lsi.upc.edu 8th June 2006 Outline Introduction
More informationSupplementary A. Built-in transition systems
L. Aufrant, G. Wisniewski PanParser (Supplementary) Supplementary A. Built-in transition systems In the following, we document all transition systems built in PanParser, along with their cost. s and s
More informationNon-Projective Dependency Parsing in Expected Linear Time
Non-Projective Dependency Parsing in Expected Linear Time Joakim Nivre Uppsala University, Department of Linguistics and Philology, SE-75126 Uppsala Växjö University, School of Mathematics and Systems
More informationTransition-based Parsing with Neural Nets
CS11-747 Neural Networks for NLP Transition-based Parsing with Neural Nets Graham Neubig Site https://phontron.com/class/nn4nlp2017/ Two Types of Linguistic Structure Dependency: focus on relations between
More informationRefresher on Dependency Syntax and the Nivre Algorithm
Refresher on Dependency yntax and Nivre Algorithm Richard Johansson 1 Introduction This document gives more details about some important topics that re discussed very quickly during lecture: dependency
More informationTransition-Based Parsing of the Chinese Treebank using a Global Discriminative Model
Transition-Based Parsing of the Chinese Treebank using a Global Discriminative Model Yue Zhang Oxford University Computing Laboratory yue.zhang@comlab.ox.ac.uk Stephen Clark Cambridge University Computer
More informationA Transition-Based Dependency Parser Using a Dynamic Parsing Strategy
A Transition-Based Dependency Parser Using a Dynamic Parsing Strategy Francesco Sartorio Department of Information Engineering University of Padua, Italy sartorio@dei.unipd.it Giorgio Satta Department
More informationBasic Parsing with Context-Free Grammars. Some slides adapted from Karl Stratos and from Chris Manning
Basic Parsing with Context-Free Grammars Some slides adapted from Karl Stratos and from Chris Manning 1 Announcements HW 2 out Midterm on 10/19 (see website). Sample ques>ons will be provided. Sign up
More informationNon-Projective Dependency Parsing with Non-Local Transitions
Non-Projective Dependency Parsing with Non-Local Transitions Daniel Fernández-González and Carlos Gómez-Rodríguez Universidade da Coruña FASTPARSE Lab, LyS Research Group, Departamento de Computación Campus
More informationDependency Parsing CMSC 723 / LING 723 / INST 725. Marine Carpuat. Fig credits: Joakim Nivre, Dan Jurafsky & James Martin
Dependency Parsing CMSC 723 / LING 723 / INST 725 Marine Carpuat Fig credits: Joakim Nivre, Dan Jurafsky & James Martin Dependency Parsing Formalizing dependency trees Transition-based dependency parsing
More informationCS224n: Natural Language Processing with Deep Learning 1 Lecture Notes: Part IV Dependency Parsing 2 Winter 2019
CS224n: Natural Language Processing with Deep Learning 1 Lecture Notes: Part IV Dependency Parsing 2 Winter 2019 1 Course Instructors: Christopher Manning, Richard Socher 2 Authors: Lisa Wang, Juhi Naik,
More informationDependency Parsing. Johan Aulin D03 Department of Computer Science Lund University, Sweden
Dependency Parsing Johan Aulin D03 Department of Computer Science Lund University, Sweden d03jau@student.lth.se Carl-Ola Boketoft ID03 Department of Computing Science Umeå University, Sweden calle_boketoft@hotmail.com
More informationDensity-Driven Cross-Lingual Transfer of Dependency Parsers
Density-Driven Cross-Lingual Transfer of Dependency Parsers Mohammad Sadegh Rasooli Michael Collins rasooli@cs.columbia.edu Presented by Owen Rambow EMNLP 2015 Motivation Availability of treebanks Accurate
More informationLearning Latent Linguistic Structure to Optimize End Tasks. David A. Smith with Jason Naradowsky and Xiaoye Tiger Wu
Learning Latent Linguistic Structure to Optimize End Tasks David A. Smith with Jason Naradowsky and Xiaoye Tiger Wu 12 October 2012 Learning Latent Linguistic Structure to Optimize End Tasks David A. Smith
More informationSorting Out Dependency Parsing
Sorting Out Dependency Parsing Joakim Nivre Uppsala University and Växjö University Sorting Out Dependency Parsing 1(38) Introduction Introduction Syntactic parsing of natural language: Who does what to
More informationTransition-based dependency parsing
Transition-based dependency parsing Syntactic analysis (5LN455) 2014-12-18 Sara Stymne Department of Linguistics and Philology Based on slides from Marco Kuhlmann Overview Arc-factored dependency parsing
More informationOnline Graph Planarisation for Synchronous Parsing of Semantic and Syntactic Dependencies
Online Graph Planarisation for Synchronous Parsing of Semantic and Syntactic Dependencies Ivan Titov University of Illinois at Urbana-Champaign James Henderson, Paola Merlo, Gabriele Musillo University
More informationAutomatic Discovery of Feature Sets for Dependency Parsing
Automatic Discovery of Feature Sets for Dependency Parsing Peter Nilsson Pierre Nugues Department of Computer Science Lund University peter.nilsson.lund@telia.com Pierre.Nugues@cs.lth.se Abstract This
More informationSorting Out Dependency Parsing
Sorting Out Dependency Parsing Joakim Nivre Uppsala University and Växjö University Sorting Out Dependency Parsing 1(38) Introduction Introduction Syntactic parsing of natural language: Who does what to
More informationCombining Discrete and Continuous Features for Deterministic Transition-based Dependency Parsing
Combining Discrete and Continuous Features for Deterministic Transition-based Dependency Parsing Meishan Zhang and Yue Zhang Singapore University of Technology and Design {meishan zhang yue zhang}@sutd.edu.sg
More informationEasy-First Chinese POS Tagging and Dependency Parsing
Easy-First Chinese POS Tagging and Dependency Parsing ABSTRACT Ji Ma, Tong Xiao, Jingbo Zhu, Feiliang Ren Natural Language Processing Laboratory Northeastern University, China majineu@outlook.com, zhujingbo@mail.neu.edu.cn,
More informationThe Application of Constraint Rules to Data-driven Parsing
The Application of Constraint Rules to Data-driven Parsing Sardar Jaf The University of Manchester jafs@cs.man.ac.uk Allan Ramsay The University of Manchester ramsaya@cs.man.ac.uk Abstract In this paper,
More informationAssignment 4 CSE 517: Natural Language Processing
Assignment 4 CSE 517: Natural Language Processing University of Washington Winter 2016 Due: March 2, 2016, 1:30 pm 1 HMMs and PCFGs Here s the definition of a PCFG given in class on 2/17: A finite set
More informationMorpho-syntactic Analysis with the Stanford CoreNLP
Morpho-syntactic Analysis with the Stanford CoreNLP Danilo Croce croce@info.uniroma2.it WmIR 2015/2016 Objectives of this tutorial Use of a Natural Language Toolkit CoreNLP toolkit Morpho-syntactic analysis
More informationMaltOptimizer: A System for MaltParser Optimization
MaltOptimizer: A System for MaltParser Optimization Miguel Ballesteros Joakim Nivre Universidad Complutense de Madrid, Spain miballes@fdi.ucm.es Uppsala University, Sweden joakim.nivre@lingfil.uu.se Abstract
More informationParsing with Dynamic Programming
CS11-747 Neural Networks for NLP Parsing with Dynamic Programming Graham Neubig Site https://phontron.com/class/nn4nlp2017/ Two Types of Linguistic Structure Dependency: focus on relations between words
More informationIncremental Integer Linear Programming for Non-projective Dependency Parsing
Incremental Integer Linear Programming for Non-projective Dependency Parsing Sebastian Riedel James Clarke ICCS, University of Edinburgh 22. July 2006 EMNLP 2006 S. Riedel, J. Clarke (ICCS, Edinburgh)
More informationManaging a Multilingual Treebank Project
Managing a Multilingual Treebank Project Milan Souček Timo Järvinen Adam LaMontagne Lionbridge Finland {milan.soucek,timo.jarvinen,adam.lamontagne}@lionbridge.com Abstract This paper describes the work
More informationA Dynamic Confusion Score for Dependency Arc Labels
A Dynamic Confusion Score for Dependency Arc Labels Sambhav Jain and Bhasha Agrawal Language Technologies Research Center IIIT-Hyderabad, India {sambhav.jain, bhasha.agrawal}@research.iiit.ac.in Abstract
More informationDependency grammar and dependency parsing
Dependency grammar and dependency parsing Syntactic analysis (5LN455) 2016-12-05 Sara Stymne Department of Linguistics and Philology Based on slides from Marco Kuhlmann Activities - dependency parsing
More informationDependency grammar and dependency parsing
Dependency grammar and dependency parsing Syntactic analysis (5LN455) 2014-12-10 Sara Stymne Department of Linguistics and Philology Based on slides from Marco Kuhlmann Mid-course evaluation Mostly positive
More informationTopics in Parsing: Context and Markovization; Dependency Parsing. COMP-599 Oct 17, 2016
Topics in Parsing: Context and Markovization; Dependency Parsing COMP-599 Oct 17, 2016 Outline Review Incorporating context Markovization Learning the context Dependency parsing Eisner s algorithm 2 Review
More informationImproving Transition-Based Dependency Parsing with Buffer Transitions
Improving Transition-Based Dependency Parsing with Buffer Transitions Daniel Fernández-González Departamento de Informática Universidade de Vigo Campus As Lagoas, 32004 Ourense, Spain danifg@uvigo.es Carlos
More informationAT&T: The Tag&Parse Approach to Semantic Parsing of Robot Spatial Commands
AT&T: The Tag&Parse Approach to Semantic Parsing of Robot Spatial Commands Svetlana Stoyanchev, Hyuckchul Jung, John Chen, Srinivas Bangalore AT&T Labs Research 1 AT&T Way Bedminster NJ 07921 {sveta,hjung,jchen,srini}@research.att.com
More informationEDAN20 Language Technology Chapter 13: Dependency Parsing
EDAN20 Language Technology http://cs.lth.se/edan20/ Pierre Nugues Lund University Pierre.Nugues@cs.lth.se http://cs.lth.se/pierre_nugues/ September 19, 2016 Pierre Nugues EDAN20 Language Technology http://cs.lth.se/edan20/
More informationOnline Service for Polish Dependency Parsing and Results Visualisation
Online Service for Polish Dependency Parsing and Results Visualisation Alina Wróblewska and Piotr Sikora Institute of Computer Science, Polish Academy of Sciences, Warsaw, Poland alina@ipipan.waw.pl,piotr.sikora@student.uw.edu.pl
More informationDependency grammar and dependency parsing
Dependency grammar and dependency parsing Syntactic analysis (5LN455) 2015-12-09 Sara Stymne Department of Linguistics and Philology Based on slides from Marco Kuhlmann Activities - dependency parsing
More informationNLP Chain. Giuseppe Castellucci Web Mining & Retrieval a.a. 2013/2014
NLP Chain Giuseppe Castellucci castellucci@ing.uniroma2.it Web Mining & Retrieval a.a. 2013/2014 Outline NLP chains RevNLT Exercise NLP chain Automatic analysis of texts At different levels Token Morphological
More informationStatistical parsing. Fei Xia Feb 27, 2009 CSE 590A
Statistical parsing Fei Xia Feb 27, 2009 CSE 590A Statistical parsing History-based models (1995-2000) Recent development (2000-present): Supervised learning: reranking and label splitting Semi-supervised
More informationDependency Parsing domain adaptation using transductive SVM
Dependency Parsing domain adaptation using transductive SVM Antonio Valerio Miceli-Barone University of Pisa, Italy / Largo B. Pontecorvo, 3, Pisa, Italy miceli@di.unipi.it Giuseppe Attardi University
More informationHomework 2: HMM, Viterbi, CRF/Perceptron
Homework 2: HMM, Viterbi, CRF/Perceptron CS 585, UMass Amherst, Fall 2015 Version: Oct5 Overview Due Tuesday, Oct 13 at midnight. Get starter code from the course website s schedule page. You should submit
More informationDependency Parsing. Ganesh Bhosale Neelamadhav G Nilesh Bhosale Pranav Jawale under the guidance of
Dependency Parsing Ganesh Bhosale - 09305034 Neelamadhav G. - 09305045 Nilesh Bhosale - 09305070 Pranav Jawale - 09307606 under the guidance of Prof. Pushpak Bhattacharyya Department of Computer Science
More informationUndirected Dependency Parsing
Computational Intelligence, Volume 59, Number 000, 2010 Undirected Dependency Parsing CARLOS GÓMEZ-RODRÍGUEZ cgomezr@udc.es Depto. de Computación, Universidade da Coruña. Facultad de Informática, Campus
More informationIncremental Joint POS Tagging and Dependency Parsing in Chinese
Incremental Joint POS Tagging and Dependency Parsing in Chinese Jun Hatori 1 Takuya Matsuzaki 1 Yusuke Miyao 3 Jun ichi Tsujii 4 1 University of Tokyo / 7-3-1 Hongo, Bunkyo, Tokyo, Japan 2 National Institute
More informationStanford s System for Parsing the English Web
Stanford s System for Parsing the English Web David McClosky s, Wanxiang Che h, Marta Recasens s, Mengqiu Wang s, Richard Socher s, and Christopher D. Manning s s Natural Language Processing Group, Stanford
More informationStructured Prediction Basics
CS11-747 Neural Networks for NLP Structured Prediction Basics Graham Neubig Site https://phontron.com/class/nn4nlp2017/ A Prediction Problem I hate this movie I love this movie very good good neutral bad
More informationForest-based Algorithms in
Forest-based Algorithms in Natural Language Processing Liang Huang overview of Ph.D. work done at Penn (and ISI, ICT) INSTITUTE OF COMPUTING TECHNOLOGY includes joint work with David Chiang, Kevin Knight,
More informationNatural Language Processing
Natural Language Processing Classification III Dan Klein UC Berkeley 1 Classification 2 Linear Models: Perceptron The perceptron algorithm Iteratively processes the training set, reacting to training errors
More informationFinal Project Discussion. Adam Meyers Montclair State University
Final Project Discussion Adam Meyers Montclair State University Summary Project Timeline Project Format Details/Examples for Different Project Types Linguistic Resource Projects: Annotation, Lexicons,...
More informationCollins and Eisner s algorithms
Collins and Eisner s algorithms Syntactic analysis (5LN455) 2015-12-14 Sara Stymne Department of Linguistics and Philology Based on slides from Marco Kuhlmann Recap: Dependency trees dobj subj det pmod
More informationExploring Automatic Feature Selection for Transition-Based Dependency Parsing
Procesamiento del Lenguaje Natural, Revista nº 51, septiembre de 2013, pp 119-126 recibido 18-04-2013 revisado 16-06-2013 aceptado 21-06-2013 Exploring Automatic Feature Selection for Transition-Based
More informationOn Structured Perceptron with Inexact Search, NAACL 2012
On Structured Perceptron with Inexact Search, NAACL 2012 John Hewitt CIS 700-006 : Structured Prediction for NLP 2017-09-23 All graphs from Huang, Fayong, and Guo (2012) unless otherwise specified. All
More informationLSA 354. Programming Assignment: A Treebank Parser
LSA 354 Programming Assignment: A Treebank Parser Due Tue, 24 July 2007, 5pm Thanks to Dan Klein for the original assignment. 1 Setup To do this assignment, you can use the Stanford teaching computer clusters.
More informationarxiv: v2 [cs.cl] 24 Mar 2015
Yara Parser: A Fast and Accurate Dependency Parser Mohammad Sadegh Rasooli 1 and Joel Tetreault 2 1 Department of Computer Science, Columbia University, New York, NY, rasooli@cs.columbia.edu 2 Yahoo Labs,
More informationHC-search for Incremental Parsing
HC-search for Incremental Parsing Yijia Liu, Wanxiang Che, Bing Qin, Ting Liu Research Center for Social Computing and Information Retrieval Harbin Institute of Technology, China {yjliu,car,qinb,tliu}@ir.hit.edu.cn
More informationContext Encoding LSTM CS224N Course Project
Context Encoding LSTM CS224N Course Project Abhinav Rastogi arastogi@stanford.edu Supervised by - Samuel R. Bowman December 7, 2015 Abstract This project uses ideas from greedy transition based parsing
More informationLarge-Scale Syntactic Processing: Parsing the Web. JHU 2009 Summer Research Workshop
Large-Scale Syntactic Processing: JHU 2009 Summer Research Workshop Intro CCG parser Tasks 2 The Team Stephen Clark (Cambridge, UK) Ann Copestake (Cambridge, UK) James Curran (Sydney, Australia) Byung-Gyu
More informationThe CKY Parsing Algorithm and PCFGs. COMP-550 Oct 12, 2017
The CKY Parsing Algorithm and PCFGs COMP-550 Oct 12, 2017 Announcements I m out of town next week: Tuesday lecture: Lexical semantics, by TA Jad Kabbara Thursday lecture: Guest lecture by Prof. Timothy
More informationGrounded Compositional Semantics for Finding and Describing Images with Sentences
Grounded Compositional Semantics for Finding and Describing Images with Sentences R. Socher, A. Karpathy, V. Le,D. Manning, A Y. Ng - 2013 Ali Gharaee 1 Alireza Keshavarzi 2 1 Department of Computational
More informationOnline Learning of Approximate Dependency Parsing Algorithms
Online Learning of Approximate Dependency Parsing Algorithms Ryan McDonald Fernando Pereira Department of Computer and Information Science University of Pennsylvania Philadelphia, PA 19104 {ryantm,pereira}@cis.upenn.edu
More informationHidden Markov Models. Natural Language Processing: Jordan Boyd-Graber. University of Colorado Boulder LECTURE 20. Adapted from material by Ray Mooney
Hidden Markov Models Natural Language Processing: Jordan Boyd-Graber University of Colorado Boulder LECTURE 20 Adapted from material by Ray Mooney Natural Language Processing: Jordan Boyd-Graber Boulder
More informationOptimistic Backtracking A Backtracking Overlay for Deterministic Incremental Parsing
Optimistic Backtracking A Backtracking Overlay for Deterministic Incremental Parsing Gisle Ytrestøl Department of Informatics University of Oslo gisley@ifi.uio.no Abstract This paper describes a backtracking
More informationStatistical Dependency Parsing
Statistical Dependency Parsing The State of the Art Joakim Nivre Uppsala University Department of Linguistics and Philology joakim.nivre@lingfil.uu.se Statistical Dependency Parsing 1(29) Introduction
More informationEnhancing applications with Cognitive APIs IBM Corporation
Enhancing applications with Cognitive APIs After you complete this section, you should understand: The Watson Developer Cloud offerings and APIs The benefits of commonly used Cognitive services 2 Watson
More informationComputationally Efficient M-Estimation of Log-Linear Structure Models
Computationally Efficient M-Estimation of Log-Linear Structure Models Noah Smith, Doug Vail, and John Lafferty School of Computer Science Carnegie Mellon University {nasmith,dvail2,lafferty}@cs.cmu.edu
More informationArmy Research Laboratory
Army Research Laboratory Arabic Natural Language Processing System Code Library by Stephen C. Tratz ARL-TN-0609 June 2014 Approved for public release; distribution is unlimited. NOTICES Disclaimers The
More informationCS 288: Statistical NLP Assignment 1: Language Modeling
CS 288: Statistical NLP Assignment 1: Language Modeling Due September 12, 2014 Collaboration Policy You are allowed to discuss the assignment with other students and collaborate on developing algorithms
More informationRube Goldberg Final Report Format
Rube Goldberg Final Report Format Group Assignment Your team is responsible for composing a single final report that describes in detail your Rube Goldberg machine. The audience of this report is the instructor
More informationDepPattern User Manual beta version. December 2008
DepPattern User Manual beta version December 2008 Contents 1 DepPattern: A Grammar Based Generator of Multilingual Parsers 1 1.1 Contributions....................................... 1 1.2 Supported Languages..................................
More informationDeliverable D1.4 Report Describing Integration Strategies and Experiments
DEEPTHOUGHT Hybrid Deep and Shallow Methods for Knowledge-Intensive Information Extraction Deliverable D1.4 Report Describing Integration Strategies and Experiments The Consortium October 2004 Report Describing
More informationMeaning Banking and Beyond
Meaning Banking and Beyond Valerio Basile Wimmics, Inria November 18, 2015 Semantics is a well-kept secret in texts, accessible only to humans. Anonymous I BEG TO DIFFER Surface Meaning Step by step analysis
More informationCOMP 412, Fall 2018 Lab 1: A Front End for ILOC
COMP 412, Lab 1: A Front End for ILOC Due date: Submit to: Friday, September 7, 2018 at 11:59 PM comp412code@rice.edu Please report suspected typographical errors to the class Piazza site. We will issue
More informationAdvanced Search Algorithms
CS11-747 Neural Networks for NLP Advanced Search Algorithms Daniel Clothiaux https://phontron.com/class/nn4nlp2017/ Why search? So far, decoding has mostly been greedy Chose the most likely output from
More informationDependency Parsing with Undirected Graphs
Dependency Parsing with Undirected Graphs Carlos Gómez-Rodríguez Departamento de Computación Universidade da Coruña Campus de Elviña, 15071 A Coruña, Spain carlos.gomez@udc.es Daniel Fernández-González
More informationThe ATILF-LLF System for Parseme Shared Task: a Transition-based Verbal Multiword Expression Tagger
The ATILF-LLF System for Parseme Shared Task: a Transition-based Verbal Multiword Expression Tagger Hazem Al Saied Université de Lorraine, ATILF, CNRS Nancy, France halsaied@atilf.fr Abstract We describe
More informationParsing the Language of Web 2.0
Parsing the Language of Web 2.0 Jennifer Foster Joint work with Joachim Wagner, Özlem Çetinoğlu, Joseph Le Roux, Joakim Nivre, Anton Bryl, Rasul Kaljahi, Johann Roturier, Deirdre Hogan, Raphael Rubino,
More informationSchool of Computing and Information Systems The University of Melbourne COMP90042 WEB SEARCH AND TEXT ANALYSIS (Semester 1, 2017)
Discussion School of Computing and Information Systems The University of Melbourne COMP9004 WEB SEARCH AND TEXT ANALYSIS (Semester, 07). What is a POS tag? Sample solutions for discussion exercises: Week
More informationERRATOR: a Tool to Help Detect Annotation Errors in the Universal Dependencies Project
ERRATOR: a Tool to Help Detect Annotation Errors in the Universal Dependencies Project Guillaume Wisniewski LIMSI, CNRS, Univ. Paris-Sud, Université Paris Saclay, 91 403 Orsay, France guillaume.wisniewski@limsi.fr
More informationBackpropagating through Structured Argmax using a SPIGOT
Backpropagating through Structured Argmax using a SPIGOT Hao Peng, Sam Thomson, Noah A. Smith @ACL July 17, 2018 Overview arg max Parser Downstream task Loss L Overview arg max Parser Downstream task Head
More informationLecture 3: Linear Classification
Lecture 3: Linear Classification Roger Grosse 1 Introduction Last week, we saw an example of a learning task called regression. There, the goal was to predict a scalar-valued target from a set of features.
More informationCreate Swift mobile apps with IBM Watson services IBM Corporation
Create Swift mobile apps with IBM Watson services Create a Watson sentiment analysis app with Swift Learning objectives In this section, you ll learn how to write a mobile app in Swift for ios and add
More informationTraining for Fast Sequential Prediction Using Dynamic Feature Selection
Training for Fast Sequential Prediction Using Dynamic Feature Selection Emma Strubell Luke Vilnis Andrew McCallum School of Computer Science University of Massachusetts, Amherst Amherst, MA 01002 {strubell,
More informationA CASE STUDY: Structure learning for Part-of-Speech Tagging. Danilo Croce WMR 2011/2012
A CAS STUDY: Structure learning for Part-of-Speech Tagging Danilo Croce WM 2011/2012 27 gennaio 2012 TASK definition One of the tasks of VALITA 2009 VALITA is an initiative devoted to the evaluation of
More information